Deflake repaint layout tests which assumed scroll would happen before first paint.

LayoutTests/fast/repaint/horizontal-bt-overflow-child-expected.txt

Index: LayoutTests/fast/repaint/horizontal-bt-overflow-child-expected.txt
diff --git a/LayoutTests/fast/repaint/horizontal-bt-overflow-child-expected.txt b/LayoutTests/fast/repaint/horizontal-bt-overflow-child-expected.txt
index 59e9a6aea93eba521131482132d0c8b5532e3fc5..383ef4019de6f94ad488ff656bf2cfce46f54c22 100644
--- a/LayoutTests/fast/repaint/horizontal-bt-overflow-child-expected.txt
+++ b/LayoutTests/fast/repaint/horizontal-bt-overflow-child-expected.txt
@@ -6,7 +6,11 @@
       "contentsOpaque": true,
       "drawsContent": true,
       "repaintRects": [
-        [29, 106, 100, 100]
+        [29, 106, 100, 100],
+        [29, 21, 285, 201],
+        [29, 21, 100, 201],
+        [29, 20, 285, 186],
+        [29, 20, 100, 86]

[Xianzhu, 2014/12/12 16:42:00]

Could you explain this change?

[rmcilroy, 2014/12/12 17:20:17]

I'm assuming these are the invalidations which are due to the scroll.  Are you
meaning the last invalidation in particular?

[Xianzhu, 2014/12/12 17:37:36]

I mean all the new invalidations. The scroll happens before the repaint test
(and before a displayAsyncThen), so invalidations caused by the scroll should
happen before we start tracking the invalidations and should not be in the
result.

Was this result generated when the scheduler was on or off? If it generates this
result when scheduler is off, I'd like to debug and see what happened.

       ]
     }
   ]